Identification and Estimation of Time-Varying Nonseparable Panel Data Models without Stayers

• DOI: 10.48550/arxiv.1712.09222
• Publication Date: 2017-01-01
• Authors: Takuya Ishihara
• Keywords: Unobservable; Identifiability; Identification; Conditional expectation

This paper explores the identification and estimation of nonseparable panel data models. We show that structural function is nonparametrically identified when it strictly increasing in a scalar unobservable variable, conditional distributions variables do not change over time, joint support explanatory satisfies some weak assumptions. To identify target parameters, existing studies assume does there are "stayers", namely individuals with same regressor values two time periods. Our approach, by contrast, allows to depend on period an arbitrary manner require existence stayers. In part paper, we consider parametric models develop estimator implements our results. then consistency asymptotic normality estimator. Monte Carlo indicate performs well finite samples. Finally, extend results discrete outcomes, partially identified.